Prosecutors are "servant[s] of the law" and should "prosecute with earnestness and vigor."

*1243 Berger v. United States , 295 U.S. 78 , 88, 55 S.Ct. 629 , 79 L.Ed. 1314 (1935). But though the prosecutor "may strike hard blows, he is not at liberty to strike foul ones." Id.

More than fifty years ago, Brady v. Maryland , 373 U.S. 83 , 87, 83 S.Ct. 1194 , 10 L.Ed.2d 215 (1963), established that a prosecutor's suppression of material evidence favorable to the accused amounts to a foul blow. An actionable Brady violation-where the government withholds evidence that reasonably probably changes the outcome of a defendant's trial-deprives the defendant of a fundamentally fair trial. Yet because of the nature of a Brady violation, a defendant, through no fault of his own, may not learn that such a violation even occurred until years after his conviction has become final and he has already filed a motion for post-conviction relief concerning other matters.

Meanwhile, the Antiterrorism and Effective Death Penalty Act ("AEDPA") imposes limitations on post-conviction relief a prisoner may obtain. This case examines whether under those limitations, a Brady claim can ever be cognizable in a second-in-time post-conviction motion under 28 U.S.C. § 2255 if it does not meet the criteria under the statute's "gatekeeping" provision, 28 U.S.C. § 2255 (h). And that presents a question of first impression in this Circuit.

But that the case involves an issue of first impression does not necessarily mean we are writing on a clean slate. As it turns out, our Circuit has already written all over this slate. Indeed, we decided this issue's fraternal twin-whether a Brady claim can ever be cognizable in a second-in-time 28 U.S.C. § 2254 petition if it does not meet any of the criteria under 28 U.S.C. § 2244 (b)(3)(A) -in Tompkins v. Secretary, Department of Corrections , 557 F.3d 1257 (11th Cir. 2009). Because we cannot distinguish Tompkins 's reasoning from the facts or law at issue here, our Circuit's prior-precedent rule binds us to apply Tompkins 's rule: a second-in-time collateral motion based on a newly revealed Brady violation is not cognizable if it does not satisfy one of AEDPA's gatekeeping criteria for second-or-successive motions.

Though we have great respect for our colleagues, we think Tompkins got it wrong: Tompkins 's rule eliminates the sole fair opportunity for these petitioners to obtain relief. In our view, Supreme Court precedent, the nature of the right at stake here (the right to a fundamentally fair trial), and the Suspension Clause of the U.S. Constitution, Art. I, § 9, cl. 2, do not allow this. Instead, they require the conclusion that a second-in-time collateral claim based on a newly revealed actionable Brady violation is not second-or-successive for purposes of AEDPA. Consequently, such a claim is cognizable, regardless of whether it meets AEDPA's second-or-successive gatekeeping criteria.

Petitioner-Appellant Gino Scott's Brady claim may or may not be an actionable Brady violation. But we think that the district court in the first instance should have the chance to address that question by determining whether Scott's Brady claim is, in fact, actionable-a question the district court never had reason to reach. Tompkins 's rule precludes this from happening because it prohibits second-in-time collateral petitions based on all types of Brady claims-actionable and inactionable, alike-simply because they are Brady claims.

Establishing the correct rule and framework for determining whether any particular second-in-time collateral motion based on a Brady claim is cognizable is critically important to maintaining the integrity of our judicial system. No conviction resulting from a fundamentally unfair trial *1244 should be permitted to stand. 1 And when a petitioner could not have reasonably been expected to discover an actionable Brady violation before filing his first federal collateral-review motion, precluding the filing of a second-in-time petition addressing the newly discovered violation is doubly wrong. It rewards the government for its unfair prosecution and condemns the petitioner for a crime that a jury in a fair trial may well have acquitted him of. This not only corrodes faith in our system of justice, but it undermines justice itself, and it cannot be allowed. So we urge the Court to rehear this case en banc to establish the rule that our Constitution and Supreme Court precedent require.

I.

In 2003, a grand jury indicted Scott and his codefendant Jose Tamayo for conspiracy to possess with intent to distribute at least five kilograms of cocaine, in violation of 21 U.S.C. §§ 841 (b)(1)(A) and 846. Tamayo pled guilty, but Scott elected to go to trial.

At trial, the government presented evidence that Scott and Tamayo, longtime friends who made occasional drug deals, agreed with each other to buy cocaine from a couple of dealers in Jacksonville, Florida. Under the plan, Scott would give Tamayo cash, and Tamayo would drive from their hometown of Valdosta, Georgia, down to Jacksonville to make the purchase. To ensure the dealers' bona fides, Scott first arranged to meet one of them before any money changed hands. But unbeknownst to Scott and Tamayo, the purported cocaine dealer he met was actually a government informant named Freddy Pena.

Pena did his job well, and Scott was satisfied. So Scott gave Tamayo $54,000 in cash to make the purchase. Tamayo then drove to Jacksonville and met Pena. No sooner did they convene than law enforcement arrived on the scene and arrested Tamayo.

Law enforcement presented Tamayo with an offer to cooperate, and he agreed. At their direction, Tamayo made several recorded phone calls to Scott in which Scott incriminated himself in the deal. Law enforcement then arrested Scott, too, charging him with conspiracy to possess cocaine for distribution.

To prove its case, among other evidence, the government called two DEA agents who showed the jury wads of $100 bills confiscated from Scott upon his arrest.

The government also presented Tamayo. He testified that he and Scott went together to the pre-purchase meetings with Pena, that Scott gave him the $54,000 to purchase the cocaine, and that after getting arrested, Tamayo made a number of recorded phone calls to Scott in which Scott made incriminating statements. The government also played recordings of those phone calls for the jury.

Besides this evidence, the government put on Pena to testify about his pre-purchase meeting with Scott. In its direct examination of Pena, the government prompted him to disclose four items of information that prosecutors had previously revealed to Scott through pretrial disclosures of evidence tending to impeach Pena, disclosures required under Giglio v. United States , 405 U.S. 150 , 92 S.Ct. 763 , 31 L.Ed.2d 104 (1972). Those four items included the following: (1) that Pena was convicted in 1996 for conspiring to distribute heroin, (2) that the DEA had paid Pena more than $168,000 for cooperation *1245 on about sixteen cases since 2001, (3) that Pena had been paid $3,500 for Scott's case so far, and (4) that Pena would likely receive additional payment in the future.

To offset any negative effect of Pena's answers to these questions, the government also asked Pena whether he had ever given testimony or information to the DEA that was "false or misleading," to which Pena replied, "No, sir." Then the government inquired as to whether Pena had told the truth in his past testimony as an informant. Pena answered, "Always."

As it turns out, Pena's answers to these questions were false. But as we explain later, many years passed before the prosecuting U.S. Attorney's Office realized that the government was in possession of information demonstrating the falsity of Pena's answers and therefore before the prosecuting U.S. Attorney's Office disclosed this information to Scott.

In the meantime, and without any knowledge of this information during the trial, on cross-examination, Scott's attorney reiterated the details of Pena's heroin-trafficking conviction and emphasized how Pena benefited from working as an informant. Pena acknowledged that he stood to receive more than $10,000 from the drug money seized from Scott. He also agreed that for him, the alternative to working as an informant would be to make ends meet through strenuous manual labor. At no point did Scott's attorney confront Pena about his past truthfulness in other cases.

In its closing argument, the prosecution acknowledged Pena's monetary motive for testifying against Scott. But the prosecution emphasized that Pena "had performed successfully for DEA in the past and they continued to use him." Scott's attorney addressed Pena only briefly, noting that Pena needed the money he received working as a government informant because the job was one of only a few career options he had as a convicted felon. The jury convicted Scott, and the district court sentenced him to life in prison.

II.

Soon after his conviction, Scott filed a direct appeal. United States v. Scott , 136 Fed.Appx. 273 (11th Cir. 2005). In his appeal, Scott raised a number of issues, including, as relevant here, a claim that his trial counsel had been ineffective for failing to conduct an adequate investigation of Pena's background. Id. at 275 . We affirmed Scott's conviction, though we declined to address his ineffective-assistance claim because the record on that issue had not been developed at that point. Id. at 275, 279 . Scott sought certiorari, and the Supreme Court denied his petition on October 17, 2005. See Scott v. United States , 546 U.S. 970 , 126 S.Ct. 502 , 163 L.Ed.2d 380 (2005).

In 2006, Scott filed his first motion to vacate under 28 U.S.C. § 2255 (the "2006 Motion"). Among other claims, Scott again argued that his trial counsel was constitutionally ineffective under Strickland v. Washington , 466 U.S. 668 , 104 S.Ct. 2052 , 80 L.Ed.2d 674 (1984), for failing to properly investigate Pena ahead of trial. The district court denied the claim. Notably, however, it concluded that even if his trial counsel did exhibit deficient performance, Scott could not show that he was prejudiced because he "fail[ed] to show what additional information could have been uncovered to further impeach the witness at trial." Scott appealed on other grounds, and we affirmed. See Scott v. United States , 325 Fed.Appx. 822 , 825 (11th Cir. 2009).

Roughly five years later, in the spring of 2011, Scott's prosecutors notified the district court of impeachment information about Pena purportedly unknown to them at the time of Scott's trial. Federal prosecutors in another jurisdiction had recently *1246 advised them of the following: (1) Pena lied to law enforcement in 1996 when he was arrested for conspiracy to distribute heroin; (2) Pena admitted to Tampa DEA agents in 2002 that he had stolen cocaine from a drug dealer the year before; and (3) as a result of his admission in 2002, though no charges were brought against him, a prosecutor at the time said he would be hesitant to use Pena in future cases, and the Tampa DEA moved Pena to "restricted use." Scott's prosecutors described their failure to include this among their required pre-trial disclosures as "inadvertent," maintaining that they were "unaware of this information until almost 7 years after the trial."

Based on this information, on November 17, 2011, Scott filed another motion under § 2255 to vacate his conviction and sentence ("2011 Motion"). In his 2011 Motion, Scott asserted for the first time that the Government had obtained his conviction by violating Brady , 373 U.S. 83 , 83 S.Ct. 1194 , and Giglio , 405 U.S. at 150 , 92 S.Ct. 763 . Both of these claims relied on the Government's 2011 disclosure of evidence relating to Pena. Scott asserted that had the government before his trial turned over the evidence disclosed in 2011, it is reasonably probable that he would not have been convicted.

In explaining how the government's failure to disclose the information affected his trial, Scott pointed to Pena's statement that he had never given "false or misleading" testimony during his time as an informant. He complained that in its closing remarks at trial, the government argued "that although Pena had been convicted of conspiracy to distribute heroin in the 1990s, he had paid his debt to society, accepted responsibility, and then moved on into this line of work that involved essentially working with DEA in 2001." Indeed, Scott emphasized, the government represented there was "no question that Pena had performed successfully for DEA in the past and they continued to use him." But based on the evidence the government disclosed in 2011, Scott argued that Pena's testimony and the government's statements at trial were false, and the government knew or should have known this at the time. Finally, Scott urged that the testimony and statements were not harmless beyond a reasonable doubt.

To explain his failure to raise these issues on direct appeal, Scott explained that he was not aware of the information at the time. And because the information was "known only to the government" as of the time of trial, and the government had assured Scott and the trial court that it had turned over all Brady material, Scott reasoned, he could not have discovered the recently disclosed information earlier through the exercise of due diligence.

The government moved to dismiss Scott's 2011 Motion, asserting it was barred as "second or successive" under 28 U.S.C. § 2255 (h). The district court agreed, concluding it was bound by our decision in Tompkins , 557 F.3d 1257 . In Tompkins , a panel of this Court held that a second-in-time habeas petition raising claims under Brady and Giglio and brought under 28 U.S.C. § 2254 always counts as "second or successive" and is therefore subject to AEDPA's gatekeeping restrictions on second or successive petitions.

Though the district court dismissed Scott's 2011 Motion, it granted Scott's alternative motion to reopen his original 2006 Motion pursuant to Federal Rule of Civil Procedure 60(b)(3), which permits a court to reopen a final judgment on various grounds, including "fraud ..., misrepresentation, or misconduct by an opposing party." The court then reevaluated Scott's 2006 Motion in light of the new information about Pena and once again denied it. In reconsidering Scott's Strickland claim *1247 in light of the newly revealed evidence, the district court concluded that Scott's trial counsel did not exhibit constitutionally deficient performance in violation of Strickland by failing to conduct further investigation of Pena. The court did not address Strickland 's prejudice prong. Scott then appealed.

III.

"In an appeal challenging a § 2255 ruling, we review legal issues de novo and factual findings for clear error." Murphy v. United States , 634 F.3d 1303 , 1306 (11th Cir. 2011). We review a district court's order on a Rule 60(b)(3) motion for abuse of discretion. Am. Bankers Ins. Co. of Fla. v. Nw. Nat'l Ins. Co. , 198 F.3d 1332 , 1338 (11th Cir. 1999).

IV.

We first address whether the district court correctly concluded that 28 U.S.C. § 2255 (h) bars Scott's 2011 Motion as "second or successive." Section 2255(h) functions as a "gatekeeping provision" for "second or successive" motions to vacate brought under AEDPA. Under section 2255(h) no "second or successive" motions may be brought unless they identify either "(1) newly discovered evidence that, if proven and viewed in light of the evidence as a whole, would be sufficient to establish by clear and convincing evidence that no reasonable factfinder would have found the movant guilty of the offense," or "(2) a new rule of constitutional law, made retroactive to cases on collateral review by the Supreme Court, that was previously unavailable." 28 U.S.C. § 2255 (h).

Neither of those exceptions applies here. So we must consider whether Scott's 2011 Motion qualifies as "second or successive." If so, we must dismiss it.

We do not get much help from AEDPA in discerning the meaning of the phrase "second or successive." In fact, AEDPA does not define the phrase. Nor is the phrase itself "self-defining." Panetti v. Quarterman , 551 U.S. 930 , 943, 127 S.Ct. 2842 , 168 L.Ed.2d 662 (2007).

But the Supreme Court has explained that "second or successive" does not capture all collateral petitions "filed second or successively in time, even when the later filings address a ... judgment already challenged in a prior ... application." 2 Id. at 944 , 127 S.Ct. 2842 . Instead, "second or successive" is a "term of art." Slack v. McDaniel , 529 U.S. 473 , 486, 120 S.Ct. 1595 , 146 L.Ed.2d 542 (2000). And since it limits the courts' jurisdiction, we read it narrowly. See Castro v. United States , 540 U.S. 375 , 381, 124 S.Ct. 786 , 157 L.Ed.2d 778 (2003) (citing Utah v. Evans , 536 U.S. 452 , 463, 122 S.Ct. 2191 , 153 L.Ed.2d 453 (2002) ).

As the Supreme Court has construed the phrase, "second or successive" "takes its full meaning from [the Supreme Court's] case law, including decisions predating the enactment of [AEDPA]." Panetti , 551 U.S. at 943-44 , 127 S.Ct. 2842 . So we must explore the relevant case law on the meaning of "second or successive."

A. Panetti v. Quarterman set forth the factors for determining whether a second-in-time petition is "second or successive."

Our starting point is the Supreme Court's decision in Panetti . In Panetti , the *1248 petitioner (named Panetti) was convicted of capital murder and sentenced to death. Id. at 937 , 127 S.Ct. 2842 . After exhausting his state-court remedies to no avail, he filed a federal petition for habeas relief under 28 U.S.C. § 2254 . It, too, was denied. Id .

The state set an execution date, and Panetti filed another state habeas claim, this time asserting for the first time that he was not mentally competent to be executed. Id. at 937-38 , 127 S.Ct. 2842 . Following the state court's denial of the petition, Panetti filed another federal habeas petition under § 2254. Id. at 938 , 127 S.Ct. 2842 . He argued that executing him while he was mentally incompetent would violate the Eighth Amendment and transgress Ford v. Wainwright , 477 U.S. 399 , 106 S.Ct. 2595 , 91 L.Ed.2d 335 (1986). See id. at 938-41, 127 S.Ct. 2842 . The district court denied his petition, and the circuit court affirmed. Id. at 941-42 , 127 S.Ct. 2842 .

The Supreme Court granted certiorari. Id . at 942, 127 S.Ct. 2842 . Before addressing the merits, the Court considered whether it had jurisdiction over Panetti's claim, in light of 28 U.S.C. § 2244 (b)(2), a habeas gatekeeping mechanism that is much like § 2255(h) but applies to federal habeas petitions seeking review of state rather than federal cases. Similar to § 2255(h), § 2244(b)(2) precludes consideration of any "claim presented in a second or successive habeas corpus application under section 2254 that was not presented in a prior application" unless it satisfies one of two exceptions-neither of which applied to Panetti's claim. 3

The Court concluded that it enjoyed jurisdiction over Panetti's case because Panetti's second-in-time § 2254 petition was not "second or successive" as that phrase is used in § 2244(b)(2) 's gatekeeping mechanism. Id. at 947 , 127 S.Ct. 2842 . In arriving at this conclusion, the Court looked solely to three considerations: (1) the implications for habeas practice if the Court found it lacked jurisdiction over Panetti's claim; (2) the purposes of AEDPA; and (3) the pre-AEDPA abuse-of-the-writ doctrine. See id . at 943-47, 127 S.Ct. 2842 .

Beginning with the implications for habeas practice, the Court first discussed the nature of a Ford claim. See id . at 943, 127 S.Ct. 2842 . Because a Ford claim asserts that a petitioner is not competent to be executed, the Court noted that such a claim does not ripen unless the petitioner both is incompetent to be executed and imminently faces execution in that state. See id . And since many years can pass between the imposition and execution of a death sentence, a petitioner may not fall into a state of mental incompetence until after the courts have resolved his first habeas petition. Id . So if "second or successive" encompassed Ford claims, a mentally competent prisoner would always have to prophylactically raise a Ford claim in his first federal habeas petition, regardless of whether he had any indication that he might eventually become incompetent, just to preserve the possibility of raising a *1249 Ford claim at a later time. Id. This practice, the Court observed, "would add to the burden imposed on courts, applicants, and the States, with no clear advantage to any." Id. at 943 , 127 S.Ct. 2842 .

On top of burdening federal habeas practice in this way, the Court concluded that treating second-in-time Ford claims as "second or successive" would also conflict with AEDPA's purposes. AEDPA was designed to "further the principles of comity, finality, and federalism." Id. at 945 , 127 S.Ct. 2842 (citation and internal quotation marks omitted). But "[a]n empty formality requiring prisoners to file unripe Ford claims neither respects the limited legal resources available to the States nor encourages the exhaustion of state remedies." Id. at 946 , 127 S.Ct. 2842 . And as for finality concerns, the Court observed they are not implicated by a Ford claim: because of the nature of a Ford claim, federal courts are generally unable to address such claims within the time frame for resolving first habeas petitions, anyway. Id.

Finally, the Court accounted for the abuse-of-the-writ doctrine, ibr.US_Case_Law.Schema.Case_Body:v1">4 id. at 947 , 127 S.Ct. 2842 , the pre-AEDPA legal doctrine "defin[ing] the circumstances in which federal courts decline to entertain a claim presented for the first time in a second or subsequent petition for a writ of habeas corpus," McCleskey v. Zant , 499 U.S. 467 , 470, 111 S.Ct. 1454 , 113 L.Ed.2d 517 (1991). Under the abuse-of-the-writ doctrine, "to determine whether an application is 'second or successive,' a court must look to the substance of the claim the application raises and decide whether the petitioner had a full and fair opportunity to raise the claim in the prior application." Magwood v. Patterson , 561 U.S. 320 , 345, 130 S.Ct. 2788 , 177 L.Ed.2d 592 (2010) (Kennedy, J., dissenting) (citing Panetti , 551 U.S. at 947 , 127 S.Ct. 2842 ). "[I]f the petitioner had no fair opportunity to raise the claim in the prior application, a subsequent application raising that claim is not 'second or successive,' and [AEDPA's] bar does not apply." Id. at 346, 130 S.Ct. 2788 (Kennedy, J., dissenting) (citing Panetti , 551 U.S. at 947 , 127 S.Ct. 2842 ). Since a Ford claim considers a petitioner's mental state at the time of proposed execution and Panetti's first § 2254 petition was filed well before that time, Panetti did not have a full and fair opportunity to raise that claim-that is, the claim did not ripen-until after his first § 2254 petition was resolved. See Panetti , 551 U.S. at 947 , 127 S.Ct. 2842 . For that reason, the Court found no abuse of the writ. Id .

So ultimately, the Supreme Court held that AEDPA's "second or successive" bar did not preclude Panetti's second-in-time petition raising a Ford claim. Id. As the Court explained, "We are hesitant to construe a statute, implemented to further the principles of comity, finality, and federalism, in a manner that would require unripe (and, often, factually unsupported) claims to be raised as a mere formality, to the benefit of no party." Id.

B. Applying the Panetti factors to an actionable Brady violation that the petitioner in exercising due diligence could not reasonably have been expected to discover in the absence of the government's disclosure yields the conclusion that such a claim is not "second or successive."

In Panetti 's light, we must consider whether second-in-time petitions raising newly disclosed actionable Brady 5 violations-where *1250 the newly disclosed evidence creates a reasonable probability that it would change the outcome of the proceeding-are "second or successive" within the meaning of § 2255(h) 's gatekeeping provision. We find that they are not. The Panetti factors and their sub-considerations uniformly require this conclusion.

1. Precluding claims based on Brady violations that a prisoner could not have discovered through due diligence would adversely affect habeas practice.

First, as the Panetti Court observed is true of Ford claims, precluding Brady claims that a prisoner could not have discovered through due diligence would adversely affect habeas practice. This is so because of the nature of a Brady claim.

Brady and its progeny stand for the proposition that the prosecution's suppression of evidence favorable to the defendant "violates due process where the evidence is material either to guilt or to punishment, irrespective of the good faith or bad faith of the prosecution." Strickler v. Greene , 527 U.S. 263 , 280, 119 S.Ct. 1936 , 144 L.Ed.2d 286 (1999) (quoting Brady , 373 U.S. at 87 , 83 S.Ct. 1194 ) (internal quotation marks omitted). Evidence is "material," in turn, when "there is a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different." Id. (citation and internal quotation marks omitted). So no actionable Brady violation occurs "unless the nondisclosure was so serious that there is a reasonable probability that the suppressed evidence would have produced a different verdict." 6 Id. at 281, 119 S.Ct. 1936 (internal quotation marks omitted).

Because of the nature of a Brady violation, the petitioner often cannot learn of such a violation at all, even when acting diligently, unless and until the government discloses it. As with second-in-time Ford claims, then, "conscientious defense attorneys would be obliged to file unripe (and, in many cases, meritless) [ Brady ] claims in each and every [first § 2255 ] application [ (and direct appeal) ]," Panetti , 551 U.S. at 943 , 127 S.Ct. 2842 , to preserve then-hypothetical claims on the chance that the government might have committed a material Brady violation that will eventually be disclosed. And also like with Ford claims, the courts would be forced to address this *1251 avalanche of substantively useless Brady claims-only there would be even more meritless Brady claims because Brady does not apply only in capital cases, like Ford does. For this reason, finding second-in-time Brady claims to be "second or successive" under § 2255 would have even more deleterious effects on habeas practice than concluding second-in-time Ford claims were "second or successive."

2. Precluding Brady claims that a petitioner could not have discovered through due diligence impedes finality interests.

Second, precluding Brady claims that a petitioner could not have discovered through due diligence actually impedes finality interests. We start from the proposition that at the very least, the second-in-time filing of a Brady claim that a prisoner could not have discovered earlier through the reasonable exercise of due diligence does not negatively implicate AEDPA's finality concerns any more than does the second-in-time filing of a Ford claim, 7 though for different reasons. To explain why, we return to the nature of a Brady violation.

When a Brady violation occurs, a defendant is entitled to a new trial. Brady , 373 U.S. at 87 , 83 S.Ct. 1194 . As the Supreme Court has explained, "[a] prosecution that withholds evidence ... which, if made available, would tend to exculpate [the defendant] or reduce the penalty[,] ... casts the prosecutor in the role of an architect of a proceeding that does not comport with standards of justice, even though ... his action is not 'the result of guile.' " Id. at 87-88 , 83 S.Ct. 1194 . Put simply, a criminal defendant does not receive a fair trial when a Brady violation occurs. 8

Yet the Constitution guarantees criminal defendants a fair trial. Delaware v. Van Arsdall , 475 U.S. 673 , 681, 106 S.Ct. 1431 , 89 L.Ed.2d 674 (1986). So imprisoning someone based on the results of an unfair trial and then precluding any remedy at all might well work a suspension of the writ of habeas corpus. Cf. Magwood , 561 U.S. at 350 , 130 S.Ct. 2788 (Kennedy, J., dissenting) (opining that refusal to consider a second-in-time habeas petition challenging an alleged violation that occurred entirely after the denial of the first petition "would be inconsistent with abuse-of-the-writ principles and might work a suspension of the writ of habeas corpus").

And even if precluding a remedy for a Brady violation that a petitioner could not reasonably have been expected to discover through due diligence does not suspend the writ, it certainly clashes with finality concerns. The Supreme Court has noted that finality is important to endow criminal law with "much of its deterrent effect." McCleskey , 499 U.S. at 491 , 111 S.Ct. 1454 (citation and quotation marks omitted). But an uncorrected unfair trial has the opposite effect.

*1252 Procedural fairness is necessary to the perceived legitimacy of the law. Kevin Burke & Steve Leben, Procedural Fairness: A Key Ingredient in Public Satisfaction , 44 Ct. Rev. 4, 7 (2007-2008) (citing Tom. R. Tyler, Psychological Perspectives on Legitimacy and Legitimation , 57 Ann. Rev. Psychol. 375 (2006) ). And legitimacy affects compliance. Cf. id. (citing studies showing reduced recidivism when defendants perceived themselves as having received fair process). When the government imprisons a person after a constitutionally unfair trial, that undermines the legitimacy of the law and its deterrent effect. A person who perceives that the government will cheat to convict him, regardless of his guilt or innocence, actually has less incentive to comply with the law because, in his view, compliance makes no difference to conviction.

But that is not the only reason that precluding second-in-time Brady claims is at odds with finality concerns. Finality is also important because giving a habeas petitioner a new trial can prejudice the government through "erosion of memory and dispersion of witnesses that occur with the passage of time." McCleskey , 499 U.S. at 491 , 111 S.Ct. 1454 (citation and internal quotation marks omitted). Yet the government alone holds the key to ensuring a Brady violation does not occur. So the government cannot be heard to complain of trial prejudice from a new trial necessitated by its own late disclosure of a Brady violation, since it is solely responsible for inflicting any such prejudice on itself in such circumstances. Whatever finality interest Congress intended for AEDPA to promote, surely it did not aim to encourage prosecutors to withhold constitutionally required evidentiary disclosures long enough that verdicts obtained as a result of government misconduct would be insulated from correction.

Finality interests then are not served by saying a prisoner has not timely brought his Brady claim where the government's failures affirmatively and entirely prevented him from doing so. Cf. Williams v. Taylor , 529 U.S. 420 , 437, 120 S.Ct. 1479 , 146 L.Ed.2d 435 (2000) (comity interests "not served by saying a prisoner 'has failed to develop the factual basis of a claim' [under § 2254(e)(2) ] where he was unable to develop his claim in state court despite diligent effort"). For this reason, finality concerns cannot justify precluding Brady claims that a prisoner could not have discovered through due diligence.

3. Precluding Brady claims that a prisoner could not have discovered through due diligence is not consistent with the abuse-of-the-writ doctrine .

Finally, allowing a second-in-time Brady claim that a prisoner could not have discovered earlier through the reasonable exercise of due diligence does not offend the abuse-of-the-writ doctrine. As we have noted, the abuse-of-the-writ doctrine calls for courts to consider whether a habeas petitioner has previously had "a full and fair opportunity to raise the claim in the prior application." Magwood , 561 U.S. at 345 , 130 S.Ct. 2788 (Kennedy, J., dissenting) (citing Panetti , 551 U.S. at 947 , 127 S.Ct. 2842 ).

To demonstrate that a petitioner has been deprived of a "full and fair opportunity," the doctrine requires him to make two showings: (1) he has "cause," or a "legitimate excuse," for failing to raise the claim earlier, McCleskey , 499 U.S. at 490 , 111 S.Ct. 1454 , and (2) he was prejudiced by the error he claims, id . at 493, 111 S.Ct. 1454 . See also Sawyer v. Whitley , 505 U.S. 333 , 338, 112 S.Ct. 2514 , 120 L.Ed.2d 269 (1992).

*1253 "Cause" explains why the petitioner could not have filed his claim earlier even "in the exercise of reasonable care and diligence." McCleskey , 499 U.S. at 493 , 111 S.Ct. 1454 . A petitioner satisfies the cause requirement where he can demonstrate "interference by officials that makes compliance with the ... procedural rule impracticable, and a showing that the factual or legal basis for a claim was not reasonably available to counsel." Id. at 493-94 , 111 S.Ct. 1454 (citation and internal quotation marks omitted). 9 A Brady violation that a prisoner could not reasonably have been expected to discover through the exercise of due diligence falls into that category. See , e.g. , Strickler , 527 U.S. at 289 , 119 S.Ct. 1936 (finding cause for failing to raise a Brady claim where the prosecution withheld exculpatory evidence, the petitioner reasonably relied on the prosecution's open-file policy, and the government asserted during state habeas proceedings "that petitioner had already received 'everything known to the government.' ").

As for prejudice, as we have noted, when a Brady violation is at issue, a petitioner must demonstrate a reasonable probability that had the government disclosed the evidence at issue, the outcome of the proceeding would have differed. Strickler , 527 U.S. at 280 , 119 S.Ct. 1936 . So a petitioner cannot establish a Brady violation without also satisfying the abuse-of-the-writ doctrine's requirement to show prejudice.

That means a petitioner can demonstrate both cause and prejudice by establishing a Brady violation that he could not reasonably have discovered through due diligence. And where a petitioner shows both cause and prejudice, he has enjoyed no "full and fair opportunity" to bring the claim earlier. To remedy this problem, the abuse-of-the-writ doctrine favors allowing such a second-in-time claim.

In short, all the Panetti factors-the implications for habeas practice, the purposes of AEDPA, and the abuse-of-the-writ doctrine-compel the conclusion that second-in-time Brady claims cannot be "second or successive" for purposes of § 2255(h). 10 And nothing Panetti teaches us to consider so much as hints otherwise.

C. Tompkins nonetheless requires us to conclude that second-in-time Brady claims are always "second or successive."

The district court, however, concluded that our decision in Tompkins v. Secretary, Department of Corrections precluded it from ruling that second-in-time Brady claims that could not have been discovered earlier through the exercise of reasonable diligence are not "second or successive." We now take a look at Tompkins to decide whether that is correct.

In Tompkins , this Court considered whether a second-in-time § 2254 petition that raised Brady and Giglio claims, *1254 among others, qualified as "second or successive" for purposes of § 2244(b)(3)(A). The Tompkins panel determined it did. 557 F.3d at 1260 .

To reach this conclusion, the panel first determined that the Supreme Court in Panetti "limit[ed] its holding to Ford claims." Tompkins , 557 F.3d at 1259 . The panel, in essence, deemed the Panetti factors irrelevant to analyzing the issue before it and further attempted to explain why Panetti was factually distinguishable from the case it was reviewing. Id. at 1260 .

1. Tompkins was incorrectly decided.

We respectfully disagree with the Tompkins panel's analysis and conclusion. As we read Panetti , the Supreme Court did not limit its analysis to petitions involving Ford claims. And when we apply the Panetti factors to Brady claims, as we must, Brady claims cannot be factually distinguished from Ford claims for purposes of determining whether they are "second or successive."

a. Panetti did not limit its analysis to petitions involving Ford claims.

Beginning with the breadth of Panetti 's holding, we cannot agree that the Supreme Court restricted its analysis to second-in-time petitions involving only Ford claims. Neither Panetti 's language nor its analysis supports such a conclusion.

First, Panetti 's language rules out such a narrow holding. In fact, the Supreme Court summarized its own jurisdictional holding as recognizing "exceptions"-plural-to the rule that a second-in-time petition fails AEDPA's "second or successive" bar: "In the usual case, a petition filed second in time and not otherwise permitted by the terms of § 2244 will not survive AEDPA's 'second or successive' bar. There are, however, exceptions ." Panetti , 551 U.S. at 947 , 127 S.Ct. 2842 (emphasis added).

Of course, that alone does not specify what exactly the Court had in mind. But the Court then immediately followed up this statement with what we understand as a partial test for determining whether a second-in-time petition that includes a particular type of claim qualifies as "second or successive": "We are hesitant to construe [AEDPA], implemented to further the principles of comity, finality, and federalism, in a manner that would require unripe (and, often, factually unsupported) claims to be raised as a mere formality, to the benefit of no party." Id. If the Court intended to limit its holding to second-in-time Ford claims only, we think it would have employed the singular form of "exception," rather than the plural, and it would have referred specifically to Ford claims in that sentence instead of stating a generally applicable rule for construing the phrase "second or successive" in AEDPA.

Second, the analysis in Panetti itself demonstrates that the Supreme Court did not limit Panetti 's holding to Ford claims. As we have noted, the Panetti Court arrived at its conclusion solely by evaluating three different generally applicable factors: the "implications for habeas practice," AEDPA's purposes, and the abuse-of-the-writ doctrine. See id. at 945-47 , 127 S.Ct. 2842 . Not one of these factors applies uniquely to Ford claims. Nor does any factor apply in such a way as to allow only Ford claims through.

Significantly, the Supreme Court also emphasized the importance of accounting for AEDPA's purposes and the implications for habeas practice not just when considering whether Ford claims are "second or successive" but whenever "petitioners run the risk under the proposed interpretation [of AEDPA] of forever losing their opportunity for any federal review of their ... claims." Id . at 945-46, 127 S.Ct. 2842 (citation and internal quotation marks *1255 omitted). Towards that end, the Court drew on examples where it had construed other aspects of AEDPA's limiting language to nonetheless allow for claims and procedures where failure to do so would preclude any opportunity for petitioners to have potentially meritorious claims heard.

For example, the Panetti Court pointed to Castro v. United States , 540 U.S. 375 , 124 S.Ct. 786 , 157 L.Ed.2d 778 (2003). In that case, the pro se petitioner filed a motion for new trial under Federal Rule of Criminal Procedure 33. Castro , 540 U.S. at 378 , 124 S.Ct. 786 . The district court recharacterized the filing as a § 2255 motion, without notice to the petitioner. The district court denied the motion on the merits, and we affirmed. Id. at 378-79 , 124 S.Ct. 786 . Three years later, when the petitioner sought to file a motion he called a § 2255 motion, the motion was dismissed as "second or successive." Id. at 379 , 124 S.Ct. 786 . The Supreme Court granted certiorari to consider whether a pro se petitioner's motion may be recharacterized as second or successive without notice to the petitioner. Id. But before the Court could consider the answer to that question, it had to determine whether it could even take up the case since § 2244(b)(3)(E) requires that the "grant or denial of an authorization by a court of appeals to file a second or successive application ... shall not be the subject of a [certiorari] petition." Id. at 379 , 124 S.Ct. 786 (quoting 28 U.S.C. § 2244 (b)(3)(E) ). The Castro Court held that it could still review the case, despite the lower courts' actions. Id.

In Panetti , the Court described its holding in Castro as having "resisted an interpretation of [AEDPA] that would produce troublesome results, create procedural anomalies, and close our doors to a class of habeas petitioners seeking review without any clear indication that such was Congress' intent." Panetti , 551 U.S. at 946 , 127 S.Ct. 2842 (internal quotation marks omitted). And the Court cited other several cases that demonstrate these same principles. See Williams , 529 U.S. at 437, 120 S.Ct. 1479 (holding that under § 2254(e)(2), a "fail[ure] to develop" a claim's factual basis in state-court proceedings is not established unless the petitioner is not duly diligent); Johnson v. United States , 544 U.S. 295 , 308-09, 125 S.Ct. 1571 , 161 L.Ed.2d 542 (2005) (holding that where an underlying state conviction used to enhance a federal sentence has since been vacated, § 2255 's one-year limitations period does not begin to run until petitioner receives notice of order vacating the prior conviction, as long as petitioner sought order with due diligence); Granberry v. Greer , 481 U.S. 129 , 131-34, 107 S.Ct. 1671 , 95 L.Ed.2d 119 (1987) (holding that where the state fails to object on grounds of exhaustion and a potentially meritorious exhaustion defense exists, a federal court should not simply dismiss the petition but should instead exercise discretion to determine whether the administration of justice would be better served by insisting on exhaustion or by instead addressing the merits of the petition); Duncan v. Walker , 533 U.S. 167 , 178, 121 S.Ct. 2120 , 150 L.Ed.2d 251 (2001) (holding that federal habeas corpus review does not toll limitation period under § 2244(d)(2) on grounds that contrary reading "would do far less to encourage exhaustion prior to seeking federal habeas review and would hold greater potential to hinder finality").

These cases involve a variety of claims and portions of AEDPA's language. But they all share one thing: to resolve each case, the Supreme Court relied on the implications for habeas practice and the purposes of AEDPA. That the Supreme Court found these considerations applicable in these different cases demonstrates definitively that Ford claims are not a one-off; rather, they are but one type of claim among several where, in construing the *1256 meaning of AEDPA's language, we must consult the implications for habeas practice and the purposes of AEDPA.

b. Brady claims are not factually distinguishable from Ford claims for the purposes of determining whether they are "second or successive."

With Panetti and its factors out of the way, Tompkins then factually distinguished Brady claims from Ford claims without applying the Panetti factors, instead creating a new test not found in Panetti . Specifically, Tompkins homed in on the Panetti Court's pronouncement that " Ford -based incompetency claims, as a general matter, are not ripe until after the time has run to file a first federal habeas petition." See Tompkins , 557 F.3d at 1259-60 (quoting Panetti , 551 U.S. at 942 , 127 S.Ct. 2842 ) (internal quotation marks omitted). Then Tompkins ascribed a meaning and significance to the term "ripe" that directly conflicts with Panetti 's analysis. In particular, Tompkins concluded that a claim's "ripeness" depends on when the violation supporting the claim occurred. Id. at 1260. And since a Brady violation happens during trial or sentencing, Tompkins reasoned, any claim based on a Brady violation necessarily ripens, at the latest, by the end of sentencing. See id .

We see two problems with this reasoning. First, the Supreme Court in Panetti did not purport to define the word "ripe." Nor does Tompkins cite anything to support its definition of the term. See id. at 1259-61. In fact, Tompkins 's definition of the word conflicts with how the term is generally understood in the law. "Ripeness" refers to "[t]he state of a dispute that has reached, but has not passed, the point when the facts have developed sufficiently to permit an intelligent and useful decision to be made." Ripeness , Black's Law Dictionary (10th ed. 2014). But when, through no fault of the petitioner, a Brady violation goes undiscovered through trial and sentencing, the facts concerning a claim based on that violation have not been developed sufficiently to permit an intelligent and useful decision to be made. Indeed, they have not been developed at all until such time as the Brady violation is discovered.

Second, and even more significantly, to the extent that Panetti referred to ripeness as a consideration within its framework for evaluating whether a second-in-time claim is "second or successive," Tompkins 's discussion of "ripeness" cannot be harmonized with Panetti 's. Panetti accounted for what it referred to as ripeness only for the purpose of evaluating the implications on habeas practice of holding an unripe claim to be "second or successive." Panetti , 551 U.S. at 943-45 , 127 S.Ct. 2842 . As we have discussed, Panetti expressed concern that holding unripe claims to be "second or successive" would flood the courts with useless claims on the off chance that such claims might later ripen. See id . at 943, 127 S.Ct. 2842 . But, of course, that is true of Brady claims that could not have been discovered earlier through due diligence. So Panetti is not distinguishable on grounds of a difference in ripeness between Ford claims and Brady claims that could not have been discovered earlier. On the contrary, Panetti 's use of ripeness in its analysis compels the conclusion that a second-in-time Brady claim that could not have been discovered earlier is not "second or successive."

2. The prior-panel-precedent rule requires us to apply Tompkins, though we are "convinced it is wrong."

Though we disagree with Tompkins and its reasoning, we recognize that it is nonetheless our precedent. Because Tompkins addresses whether Brady claims in § 2254 petitions can ever avoid being "second or successive," we must consider whether *1257 Tompkins controls the outcome when § 2255 petitions are involved. We conclude that it does.

The prior-panel-precedent rule requires subsequent panels of the court to follow the precedent of the first panel to address the relevant issue, "unless and until the first panel's holding is overruled by the Court sitting en banc or by the Supreme Court." Smith v. GTE Corp. , 236 F.3d 1292 , 1300 n.8 (11th Cir. 2001). Even when a later panel is "convinced [the earlier panel] is wrong," the later panel must faithfully follow the first panel's ruling. United States v. Steele , 147 F.3d 1316 , 1317-18 (11th Cir. 1998) (en banc). We of course are not bound by anything that is mere dictum. See Lebron v. Sec'y of Fla. Dep't of Children & Families , 772 F.3d 1352 , 1360 (11th Cir. 2014) ("[D]iscussion in dicta 'is neither the law of the case nor binding precedent.' ") (citation omitted). But our case law reflects that under the prior-panel-precedent rule, we must follow the reasoning behind a prior holding if we cannot distinguish the facts or law of the case under consideration. See Smith , 236 F.3d at 1301-04 . So we consider whether we may limit Tompkins 's holding to only Brady claims arising under § 2254.

Important differences between § 2254 and § 2255 do exist. Among others, § 2254 vindicates the concerns of comity and federalism by restricting when federal courts can reopen state criminal convictions, while § 2255, which deals with federal criminal convictions, does not.

Nor is the interest of finality exactly the same for § 2254 and § 2255 claims. "Finality has special importance in the context of a federal attack on a state conviction." McCleskey , 499 U.S. at 491 , 111 S.Ct. 1454 .

And separation-of-powers considerations drive § 2255 claims. See Bousley v. United States , 523 U.S. 614 , 620-21, 118 S.Ct. 1604 , 140 L.Ed.2d 828 (1998) (characterizing separation-of-powers concerns as "the doctrinal underpinnings of habeas review" of federal convictions and sentences); see also Welch v. United States , --- U.S. ----, 136 S.Ct. 1257 , 1268, 194 L.Ed.2d 387 (2016). But they have no relevance to § 2254 claims.

Plus, the federal government has a distinctive concern for ensuring that federal prosecutors have acted appropriately when it reviews § 2255 claims: "the United States Attorney is the representative not of an ordinary party to a controversy, but of a sovereignty whose obligation to govern impartially is as compelling as its obligation to govern at all; and whose interest, therefore, in a criminal prosecution is not that it shall win a case, but that justice shall be done." Strickler , 527 U.S. at 281 , 119 S.Ct. 1936 (internal quotation marks omitted) (quoting Berger , 295 U.S. at 88 , 55 S.Ct. 629 ).

Even the language of the two statutes' respective gatekeeping provisions differs. Compare 28 U.S.C. § 2244 (b)(2)(B) (restricting habeas review of state convictions to, among others, cases where "the factual predicate for the claim could not have been discovered previously through the exercise of due diligence"), with 28 U.S.C. § 2255 (h) (limiting federal habeas review to cases of "newly discovered evidence," among others). But see Gonzalez v. Sec., Dep't of Corr. , 366 F.3d 1253 , 1262 (11th Cir. 2004) (en banc) (recognizing "no material difference in the relevant statutory language" between gatekeeping provisions of sections 2244 and 2255 ).

All of these differences provide good reason to treat § 2254 and § 2255 claims differently under appropriate circumstances. But none of them allows us to sufficiently distinguish Tompkins 's reasoning in analyzing Brady claims under § 2254 from how we must analyze Brady claims in this Circuit under § 2255.

*1258 As we have noted, Tompkins based its determination that all Brady claims are necessarily "second or successive" on its conclusion that all Brady claims ripen during trial or, at the latest, sentencing. We have already explained why, were we starting our analysis from scratch, we would conclude that is not correct.

But we see no basis that allows us to distinguish between state and federal proceedings in this regard; Brady claims in state proceedings do not "ripen" any sooner than do Brady claims in federal proceedings under Tompkins 's definition of the word. And while federal courts have a special interest in ensuring the integrity of federal proceedings, we do not think that that fact alone explains why Brady claims in state proceedings should be treated any differently than Brady claims in federal proceedings.

For these reasons, we must conclude that Tompkins 's reasoning governs all second-in-time Brady claims, regardless of whether they are brought under § 2254 or § 2255. Despite Tompkins 's failure to adhere to-or even to attempt to apply-the Panetti factors, we must nonetheless hew to Tompkins 's command and deem Scott's 2011 Motion "second or successive" under § 2255(h). Because Tompkins is fatally flawed, however, we respectfully urge the Court to take this case en banc so we can reconsider Tompkins 's reasoning.

V.

Having concluded we must dismiss Scott's § 2255 motion as "second or successive," we now turn to Scott's alternative motion to reopen his original 2006 Motion under Rule 60(b)(3). As we noted at the outset, the district court ultimately granted Scott's alternative motion to reopen but declined to grant him relief on the merits. On appeal, neither party disputes that the district court was within its power to reopen the 2006 Motion. Scott argues, however, that the district court incorrectly concluded that he failed to adequately allege ineffective assistance of his trial counsel in light of the government's previously undisclosed evidence about Pena.

The Sixth Amendment right to counsel "is the right to effective assistance of counsel." Strickland , 466 U.S. at 686 , 104 S.Ct. 2052 (citation and internal quotation marks omitted). A claim of ineffective assistance of counsel requires a two-pronged showing: that counsel's performance was constitutionally deficient and that counsel's deficiencies prejudiced the proceeding's outcome. Id. at 693 , 104 S.Ct. 2052 . The district court concluded that even in light of the new evidence about Pena, Scott's trial counsel did not exhibit constitutionally deficient performance.

An attorney's performance fails to meet the constitutional minimum when it falls "below an objective standard of reasonableness ..., which means that it is outside the wide range of professionally competent assistance." Payne v. Allen , 539 F.3d 1297 , 1315 (11th Cir. 2008) (citations and internal quotation marks omitted). We have observed that "omissions are inevitable" because "trial lawyers, in every case, could have done something more or something different." Chandler v. United States , 218 F.3d 1305 , 1313 (11th Cir. 2000). We therefore "conduct a highly deferential review of counsel's performance and indulge the strong presumption that counsel's performance was reasonable and that counsel made all significant decisions in the exercise of reasonable professional judgment." Payne , 539 F.3d at 1315 (alteration and internal quotation marks omitted). "[T]rial counsel has not performed deficiently when a reasonable lawyer could have decided, under the circumstances, not to investigate or present particular evidence." Id. at 1316 (quoting *1259 Grayson v. Thompson , 257 F.3d 1194 , 1225 (11th Cir. 2001) ).

The district court acknowledged that Scott's counsel took the government at its word that it had produced all Brady and Giglio material, and, as a result, that his counsel did not undertake additional steps to seek further impeachment material for Pena. But the court refused to find "that no competent lawyer would have declined to expend further time and resources" on searching for Brady and Giglio material when defense counsel is "entitled presume that the government had disclosed all such matters." Scott argues on appeal that this is incorrect, and that under the district court's reasoning, "no counsel could ever be found ineffective, entitled as counsel would be to blindly rely on the presumption that the prosecution has provided the defense with all the exculpatory or impeachment material that is to be found in the case."

We conclude the district court did not abuse its discretion in declining to find Scott's trial counsel ineffective. The decision to refrain from additional investigation into Pena's background was within the "wide range of professionally competent assistance," given the inevitable choices defense lawyers must make about how to deploy their limited time and resources. See Strickland , 466 U.S. at 690 , 104 S.Ct. 2052 . An attorney's performance is not deficient in hindsight just because he or she made one choice versus another. Cf. Willis v. Newsome , 771 F.2d 1445 , 1447 (11th Cir. 1985) ("Tactical decisions do not render assistance ineffective merely because in retrospect it is apparent that counsel chose the wrong course.").

This is not to say that no attorney could ever be found ineffective for taking the government's word as grounds for refraining from further investigation. In some cases obvious red flags might exist calling for further inquiry, even where the government has assured defense counsel that it has disclosed all Brady and Giglio material. An attorney who does not investigate under those circumstances might indeed be constitutionally ineffective. But on the facts of this case, no such red flags existed. We conclude that the district court did not abuse its discretion in declining to grant Scott relief on his reopened 2006 Motion.

VI.

Ultimately, Tompkins binds us to conclude that in § 2255 cases, all second-in-time Brady claims are "second or successive" under § 2255(h), even if the petitioner could not reasonably have been expected to discover the Brady violation and there is a reasonable probability that timely disclosure of the suppressed evidence would have resulted in an acquittal. We think this conclusion conflicts with Panetti and effects a suspension of the writ of habeas corpus as it pertains to this narrow subset of Brady claims. Supreme Court precedent, the nature of the right at stake here, and habeas corpus require a petitioner who has reasonably probably been convicted because the government failed to disclose material exculpatory evidence, to have a full and fair opportunity to obtain relief. For this reason, we urge our colleagues to rehear this case en banc and reevaluate the framework we established in Tompkins .

AFFIRMED.